Automatic segmentation of perifalcine structures in two-dimensional ultrasound images using object-based image analysis

Abstract

In this paper we present an automatic segmentation method for structures of the perifalcine region in two- dimensional ultrasound (US) images. The segmentation method is implemented in context of improving image guidance during brain tumor (glioma) resection. Tissue removal and movement of brain tissue in glioma surgery considerably decrease the accuracy of neuronavigation systems which provide image guidance based on preop- erative MRI (preMRI). The scanned central cerebral structures of the perifalcine region (such as the falx, the tentorium, the corpus callosum, and adjacent gyri) in intraoperative US (iUS) can be used as guiding frame to retrieve additional, spatially information and insights into local image and tissue deformation within the preMRI scan. Therefore, a segmentation of the perifalcine structures followed by a registration with the preMRI scan in future works may be used for an intraoperative update and even a partial fusion of the ROI in both modalities during different phases of the cerebral tumor resection. The method was evaluated on 50 US images and achieved on average a Dice coefficient of 0.68, a Hausdorff distance of 2.58 mm and a Jaccard index of 0.53